PPI-IRO: a two-stage method for protein-protein interaction extraction based on interaction relation ontology

Area of Science:

• Biomedical Informatics
• Computational Biology
• Bioinformatics

Background:

• Identifying protein-protein interactions (PPIs) is crucial for understanding biological regulatory networks.
• Biomedical literature is a rapidly expanding resource for PPI data.
• Existing methods for mining PPIs require refinement for improved accuracy.

Purpose of the Study:

• To develop a novel, two-stage method for extracting protein-protein interactions (PPIs) from biomedical literature.
• To introduce the Interaction Relation Ontology (IRO) for organizing and specifying protein interaction relationships.
• To enhance the accuracy and efficiency of PPI extraction using a structured ontological approach.

Main Methods:

• A two-stage PPI extraction pipeline was developed.
• The Interaction Relation Ontology (IRO) was utilized in a binary classifier to identify sentences containing protein interactions.
• Sentence dependency parse trees were employed, guided by IRO, for precise PPI extraction.

Main Results:

• The IRO-based method demonstrated significant performance in relation sentence classification and PPI extraction.
• The PPI extraction method achieved recall rates of approximately 80% and 90% on the AIMed and BioInfer corpora, respectively.
• The method yielded F1 scores of around 54% and 66% on the AIMed and BioInfer corpora, outperforming most existing methods.

Conclusions:

• The proposed Interaction Relation Ontology (IRO) offers a robust framework for mining protein-protein interactions.
• This novel two-stage extraction method significantly enhances the accuracy of PPI identification from biomedical texts.
• The superior performance validates the effectiveness of IRO in advancing biological network analysis.